Benqi Staked Avax
Tier 2Security Analysis Report
TVL
$200.55M
Sections
7
report sections
Citations
364
evidence references
Version
v2
10/30/2026
Moderate Risk
Safety Score · out of 10
Executive Summary
Executive Summary: Benqi Staked Avax
Benqi Staked Avax is a single-product, single-chain liquid staking pool on Avalanche, and no hack is recorded for it. Its weak point is control: a few keys and multisigs can upgrade the contract, pause it or move pooled AVAX instantly, and no timelock was found. The contract logic has been reviewed by two audit firms, but the audited code was not compared with the deployed code. Confidence is good on the on-chain control facts and weaker on liquidity depth, signer identities and off-chain custody.
Safety Rating
Overall Score: 4.9/10 Dimension scores: Code security 6.0 · Economic security 3.8 · Team transparency 4.0 · Governance 1.6 · Track record 8.6
- Code security
- 6.0 points
- Economic security
- 3.8 points
- Team transparency
- 4.0 points
- Governance
- 1.6 points
- Track record
- 8.6 points
The weak scores come from governance and economics, not from known coding flaws.
Scores by Dimension
| Dimension | Key Finding |
|---|---|
| Code Security | Two firms and formal verification reviewed StakedAvax, but the audited revision is not tied to the deployed code. |
| Economic Security | Custody and control are the main weakness; no insurance fund, slashing rule or proof of reserves is shown. |
| Team Transparency | Official documents name no team members, and weak sources give conflicting founder lists. |
| Governance | Multisigs and single keys act with zero delay; the signers of the 4-of-6 upgrade Safe are not identified by any source. |
| Track Record | No hack is recorded for the product or its family; a May 2026 event in a sibling lending market is covered only by weak sources. |
What This Protocol Does
BENQI Staked AVAX lets users deposit AVAX, the Avalanche network's native coin, and receive sAVAX in return. sAVAX is a "liquid staking" receipt token: the AVAX is delegated to validators, and staking rewards raise the value of each sAVAX over time. The DefiLlama record covers this one product on Avalanche, with TVL of about USD 200.5M. BENQI lending markets, Ignite and the QI token are other products and are context only.
Documented vs On-Chain
- A BENQI post in an Aave governance thread (2022-05-03; source reliability: community publication) says the sAVAX token "is currently held in a 4/7 multisig by the team" and that the multisig would expand to 10 signers. On Avalanche at block 98915447 (2026-10-30) the upgrade Safe is 4-of-6. Whether any current signer is outside the team is not confirmed.
- The published
StakedAvax.soldocumentation copy is older than the deployed contract. It hasaccrueRewards(uint amount), while the deployed contract at block 98915447 showsaccrueRewards(). Mechanics taken only from that copy are not confirmed for the deployed version. - The Immunefi bounty scope names the rewards contract "sAVAX reward timelock". On-chain at block 98915447 (2026-10-30) its proxy admin is owned by a single address with no code, and the upgrade delay is 0. The contract's own internal delay was not measured.
- A 2022 BENQI post in the same Aave thread (source reliability: community publication) said the delegating account was "in the process of being transitioned" to an MPC (multi-party computation) model. On-chain at block 98915447 the withdraw role is held by one address with no code, and whether it is MPC-controlled is not documented.
Key Strengths
- Audit review of
StakedAvax(sAVAX product only): Halborn's four critical findings are marked fixed in the ledger. Seven of the nine audit files sit on a host not verified as auditor-controlled, which lowers confidence in claims that rest only on them. Audits of lending, Ignite and oracle products are not counted for sAVAX. - Across all nine audit files (several cover other BENQI products), only two finding groups are critical, high or medium and not fixed. For
sAVAX, all critical and high findings are fixed except the Certora admin-trust item described under Critical Findings. The other open group is an acknowledged medium on the BENQI lending protocol, a different product. - The 15-day cooldown and 2-day redeem window are confirmed on-chain at block 98915447 and match the documentation.
- An Immunefi bug bounty is live. Sourcify, a code-matching provider, reports a runtime match for the proxy. This is not independent assurance of the code or of audit-to-deployment equivalence. The bounty excludes centralization and privileged-address impacts, so it does not cover the admin-trust risk below.
- The product is live since 2022-02-17 with no record in the hacks dataset.
Key Risks
- Key or signer compromise at the custody and upgrade layer (all
sAVAXholders). One address with no code holdswithdraw, and a 4-of-6 Safe can upgrade the contract, both with a 0-second delay. Magnitude: a withdraw-key loss exposes the AVAX in the contract at that moment, which was not measured ; an upgrade-Safe compromise reaches the same balance plus later deposits, so exposure grows with deposits . The sources do not show that an upgrade alone reaches delegated AVAX. The pool today is 24,097,137 AVAX, about USD 200.5M of TVL in the current snapshot. It is not established exposure, and it bounds today's pool, not future deposits. Likelihood: low. - Frozen or slowed exit with a possible depeg and possible lending-market liquidations (
sAVAXholders and anyone using it as collateral). Redemption takes 15 days plus a 2-day window, and the Safe can change both instantly. Magnitude: temporary loss of liquidity for the whole pool ; realised losses fall on sellers, and on borrowers only if their positions are liquidated . Whether liquidations follow depends on borrower positions and market rules, which are not established ; the size of a discount is unknown because no depeg history is measured . Likelihood: low to medium for a temporary discount in a sharp AVAX sell-off , lower for the liquidation part because borrower positions and market rules are unknown , and low for a deliberate multi-week freeze, since it needs signer misconduct . No depeg or liquidation event is measured. The pool-depth and holder figures behind this come from an unverified publisher (source reliability: unverified publisher). - Backing and exchange rate that holders cannot verify (all
sAVAXholders). If delegated AVAX is lost, the rate can stay unchanged until the gap shows at redemption. Magnitude: an unknown share of pooled AVAX, because the delegated share is not quantified. Likelihood: low. - Upgrade-Safe takeover inside the exit window. A holder who has not yet completed the 15-day cooldown cannot redeem through the protocol before an instant upgrade takes effect, and their only faster exit is an AMM sale. A holder already inside the 2-day redeem window could still redeem first, if the contract holds enough liquid AVAX, and the size of that group is not measured. The cross-analysis rates the likelihood low and the impact severe; delegated P-Chain AVAX is not shown to be reachable by an upgrade alone.
- Rewards contract that feeds the exchange rate has weaker upgrade control than the Safe. Its upgrade control is a single code-less address with a 0-second delay, and the contract's own delay was not measured.
Critical Findings
- No unresolved critical finding is reported for
sAVAX; the only open critical-or-high item is the one below. - One open item is rated High (estimated), because the audit states no rating: the Certora audit (2022) treats the admin role as fully trusted, with unrestricted deposit and withdraw. The report calls it the largest code-level trust risk in the product.
- Deployed code vs audits: the comparison was not performed ("not compared"), because the exact audited revisions are not public. This is an evidence gap, not a finding of changed or unaudited code.
Recommendation
This protocol suits risk-tolerant users who accept trust in a small group of operators and signers, hold AVAX exposure anyway, and can wait out a 15-day exit. Users with low risk tolerance, or those who plan to use sAVAX as leveraged collateral, should wait for evidence of a timelock, named signers and independent proof of the staked balance. The reasons are that control sits with instant-acting keys and multisigs and that custody is off-chain and unverifiable, with no recorded loss to date.
Report Sections
- Protocol Overview
- Code Security
- Economic Security
- Social & Team
- Governance
- Cross-Risk Analysis
52 citations · click markers to view evidence
Protocol Overview
Protocol Overview: Benqi Staked Avax
What It Does
BENQI Staked AVAX lets users deposit AVAX, the Avalanche network's native coin, and receive sAVAX in return. sAVAX is a "liquid staking" receipt token. The deposited AVAX is delegated to Avalanche validators (nodes that secure the network), and staking rewards raise the value of each sAVAX over time. Users can hold sAVAX, sell it on an exchange, or use it elsewhere in DeFi while still earning staking rewards. The protocol is live on Avalanche only, and the DefiLlama record covers this one product, the sAVAX token and its staking contract.
Assessment: the product is a single-chain, admin-operated liquid staking pool. On-chain, one StakedAvax proxy contract is controlled by a 4-of-6 multisig with no measured delay, and a single externally owned account (EOA, a plain key-controlled wallet) can pull AVAX out for delegation. Users depend on BENQI's operators and off-chain systems, not only on the smart contract. The basis is the control data and the risk page cited below. Evidence is good for the current on-chain state and weaker for audit coverage of the deployed code.
- Category: liquid staking.
- Target users: AVAX holders who want staking yield without locking their funds, and DeFi users who want a staked-AVAX collateral or trading asset.
- Related BENQI products outside this record are the BENQI lending markets (a separate DefiLlama entry in the same family), Ignite and Node Voting. They are context only here, except where they share keys, as noted under Technical Architecture.
How It Works
The protocol runs on a pooled share model: users deposit AVAX into one contract, get a pro-rata claim on the pool, and the pool grows as rewards are booked.
User flow:
- A user sends AVAX to
StakedAvaxon the Avalanche C-Chain (the smart-contract chain) by callingsubmit(), and receivessAVAX. The contract mints shares, so eachsAVAXis a share of the whole pool. (source reliability: unverified publisher) - An address holding
ROLE_WITHDRAWtakes AVAX out of the contract "for delegation" withwithdraw. The AVAX is delegated to validators on the Avalanche P-Chain (the staking chain). Delegation is split between an open pool (65%) of BENQI-selected validators and a pool (35%) chosen by BENQI Miles holders. - Staking rewards accrue on the P-Chain. An address holding
ROLE_ACCRUE_REWARDSsends the rewards to the contract by callingaccrueRewards()with AVAX attached. The contract pays the protocol share to a recipient and adds the rest tototalPooledAvax. This raises the AVAX eachsAVAXis worth. (source reliability: unverified publisher) - To exit through the protocol, the user calls
requestUnlock. The user waits out a cooldown, then redeems within a short window. (source reliability: community publication) The contract burns thesAVAXon redemption, and users cannot burn it directly. If the user misses the window, the process has to be started again. (source reliability: unverified publisher) TheredeemOverdueSharesfunction handles shares held in custody for overdue requests. (source reliability: unverified publisher) - For instant exit, the user can swap
sAVAXfor AVAX on an exchange (AMM), though AMM fees may apply. - The user can instead use
sAVAXin BENQI Markets or other DeFi platforms.
- (actor)
- (contract)
- (role)
- (external)
- (role)
- (external)
- step 1flow
- step 2flow
- step 2flow
- step 3flow
- step 4flow
- step 5flow
Core mechanism: sAVAX is a share token. The AVAX per share is totalPooledAvax divided by totalShares. Rewards enter the pool only when a privileged role sends AVAX to the contract; the contract does not read P-Chain balances. At Avalanche block 98915447 (2026-10-30) the contract reported 24,097,137 AVAX of total pooled AVAX against 18,641,543 shares, about 1.29 AVAX per sAVAX. This is an accounting total, not the contract's liquid balance. The available sources do not measure how much of it sits in the contract and how much is delegated to validators. The contract counts 64,761 stakers at the same block.
Measured on-chain at that block: the cooldown is 1,296,000 seconds (15 days) and the redeem window is 172,800 seconds (2 days). This matches the documented 15 days plus 2 days.
Key smart contracts and roles (Avalanche C-Chain):
| Contract | Address | Role |
|---|---|---|
StakedAvax (sAVAX proxy) | 0x2b2c81e08f1af8835a78bb2a90ae924ace0ea4be | Token, deposits, unlock and redeem queue, exchange rate |
StakedAvax implementation | 0xb791c7a42fd0d10f90deaa906a8735f79719fa53 | Current logic behind the proxy; Sourcify reports an exact runtime and creation match |
| sAVAX reward timelock (rewards contract) | 0x09c1e991870cbc01009a4b49397a4f2a127d3784 | Holds ROLE_ACCRUE_REWARDS; named "sAVAX reward timelock" in the bug-bounty scope |
The proxy and implementation addresses come from the on-chain read and the contracts list. The bug-bounty scope names the rewards contract. The rewards contract's own timelock delay was not measured.
Yield Sources
Yield comes from Avalanche validator staking rewards, not from token emissions; sAVAX holders see it as a rising exchange rate.
- Source of yield: the AVAX is delegated to validators and "continues to earn rewards on the P-Chain", and
sAVAXvalue accrues as rewards accumulate. - Protocol cut: the deployed
StakedAvaxcode pays aprotocolRewardShareof each reward deposit to a recipient, and the storage comment defines the scale as "0-1e18 for 0-100 %". The measured value is100000000000000000, which is 10%, and the recipient is a 3-of-9 Safe. DefiLlama revenue is also about 10% of fees: 30-day fees of 1,029,516 and revenue of 102,955. - Historical rate: BENQI's April 2022 Aave listing proposal said the delegation generates "approx 7.20% APR". That is a 2022 figure from the protocol's own team in a forum post (source reliability: community publication), so it is unconfirmed and not a current rate. A web check on 2026-10-30 found no current rate published by BENQI: the BENQI app page returned an error, and the BENQI docs say the app shows the current
sAVAXAPR without giving a number. The best dated third-party reading is a DefiLlama pool page, captured on 2026-10-30, showing a 30-day average supply APY of 5.40% (source reliability: unverified publisher). That page does not state the end date of the averaging window. A DefiLlama protocol page shows a conflicting average supply APY of 24.14% that the sources do not explain. The 5.40% figure is therefore an indicative market reading, not a confirmed rate. - Sustainability: the yield depends on Avalanche network staking rewards and on validators performing. No QI-token emission to
sAVAXholders is documented in the sources. Extra yield exists only if the user deployssAVAXelsewhere, for example in BENQI Markets, where rates are set by that market.
Dependencies & Integrations
The protocol depends on Avalanche, BENQI's off-chain operators and the markets where sAVAX trades.
- Chain: Avalanche only. DefiLlama lists a single chain. No inactive-chain deployments are listed.
- Cross-chain movement and MPC: moving AVAX between the C-Chain and P-Chain uses an off-chain multi-party computation (MPC) system. A threshold of independent nodes must co-sign each transfer. If too many nodes go offline, staking, unstaking and transfers pause.
- Validators: delegation goes to validators chosen by BENQI or by BENQI Miles voters.
- Staking bot: one document says unlocks rely on a staking bot that reduces delegated AVAX to cover redemptions. This is unconfirmed, because the document is an April 2022 file from an unverified publisher and no stronger source corroborates it. (source reliability: unverified publisher)
- Oracle: BENQI documents a Chainlink sAVAX/USD price feed on Avalanche for integrators. For Aave V3, the audited
sAVAXOracleAdapterbuilds a USD price from two parts: an AVAX/USD aggregator and thesAVAX/AVAX pool ratio read from the staking contract. The sources do not show how other integrators pricesAVAX. - Integrations: BENQI Markets (the lending product) and Aave V3. An Aave V3 listing of
sAVAXwas proposed in April 2022. (source reliability: community publication) An audit covered thesAVAXoracle adapter and the listing steward for it. (source reliability: unverified publisher) The sources do not confirm the current state of that listing. - DEX liquidity: in April 2022 BENQI said
sAVAXhad more than $120m of liquidity, with Platypus at $63m and Trader Joe at $39m. (source reliability: community publication) Current liquidity depth is not available, because the exchange-depth dataset returned no candidates.
TVL & Adoption
DefiLlama's latest captured TVL is about $200.5m, all on Avalanche, and it has fallen from a 2025 peak.
- Snapshot: TVL $200,546,865.86, collected 2026-10-30.
- Chain split: 100% Avalanche.
- Trend (DefiLlama monthly series): the highest month-start value shown is $527,978,603 on 2025-10-01. The value on 2026-10-01 was $253,135,468. TVL is USD-denominated, so AVAX price moves it. DefiLlama's 30-day flow split shows a $35.8m USD fall made up of a $41.5m price effect and about $5.7m of net quantity inflow valued at current prices.
- Users: 64,761 stakers at block 98915447.
- Fees: DefiLlama shows $12.6m of fees over 365 days and $1.03m over 30 days.
- 2025-10-01
- 528.0 USD m
- 2026-10-01
- 253.1 USD m
- 2026-10-30 snapshot
- 200.5 USD m
Technical Architecture
Upgrade, pause and parameter powers sit with a few multisigs (shared-key wallets that need several signers) with no measured time delay, plus two single-key accounts. All values below are read from Avalanche C-Chain at block 98915447 (2026-10-30).
Upgradeability: StakedAvax is an EIP-1967 proxy, so BENQI can swap its logic. BENQI's risk page states that the liquid staking contracts are upgradeable.
Who controls what (measured):
| Power | Holder | Delay |
|---|---|---|
Upgrade StakedAvax | Safe 0x832c…8cd (4 of 6 signers) via proxy admin 0x2295…a66 | 0 seconds measured |
DEFAULT_ADMIN_ROLE: set cooldown, redeem period, reward share and its recipient; ROLE_SET_TOTAL_POOLED_AVAX_CAP | Same Safe 0x832c…8cd | 0 seconds |
withdraw(uint256) (move pooled AVAX out) and deposit() | One EOA 0x565f…e83 (no contract code) | 0 seconds |
pause() | Pause Guardian contract 0x10f3…2e6 (owned by Safe 0xb952…db3, 3 of 6), Safe 0x30d6…159 (2 of 13), Safe 0x832c…8cd (4 of 6) | 0 seconds |
pauseMinting() | Pause Guardian 0x10f3…2e6 (owned by Safe 0xb952…db3, 3 of 6) and Safe 0x30d6…159 (2 of 13) | 0 seconds |
resume() and resumeMinting() | Safe 0x832c…8cd (4 of 6) | 0 seconds |
accrueRewards() | Rewards contract 0x09c1…784, Safe 0x832c…8cd | 0 seconds |
Upgrade the rewards contract 0x09c1…784 | One EOA 0x0df1…679, via an admin contract it owns | 0 seconds |
Evidence for the main rows:
- The Safe
0x832c…has 6 owners and threshold 4, no modules and no guard. - The
StakedAvaxupgrade holder is that Safe, with fastest delay 0. - The
withdrawholder is one EOA. - The rewards contract's upgrade holder is one EOA.
- The
pause()holders include the 2-of-13 Safe0x30d6…. - The
setTotalPooledAvaxCapholder is the 4-of-6 Safe, and the cap now equals the maximum 256-bit integer, which is effectively no cap. - The Pause Guardian's owner is the 3-of-6 Safe
0xb952…db3. - The sAVAX pause signers
0x30d6…also control pause functions on the BENQI lending-market pause contract at0xaad2…cb7. That Safe is a shared key between this product and the lending markets.
Role-admin limits: who can grant or revoke roles (grantRole, revokeRole) was not read, so that control is not confirmed. Whether the rewards contract enforces any delay is also not measured.
Emergency mechanisms: pause() and pauseMinting() exist, and the contract was not paused at the read block. The risk page warns that emergency functions can themselves limit withdrawals and redemption. The risk page says parameters may be updated through a multisig framework.
- (role)
- (role)
- (contract)
- (role)
- (contract)
- (role)
- control
- control
- controlrisk
- flow
- controlrisk
Key Risks (Protocol-Level)
User funds are exposed mainly to operator and off-chain risk, because pooled AVAX leaves the contract for delegation under a single-key role and cross-chain MPC signing.
- Where funds sit: deposited AVAX first sits in
StakedAvaxon the C-Chain, then moves out to the P-Chain for delegation. A single EOA holds thewithdrawrole with no measured delay. Assessment: this is the most direct path to loss from key compromise or operator error; the basis is the measured role holder above plus the risk page's statement that signer compromise could "redirect staked AVAX". - Reward size is set by operators: the role holder decides how much AVAX to send to
accrueRewards()and when. The contract does not compare it with P-Chain balances, so the exchange rate trusts the rewards contract and the 4-of-6 Safe. Under-reporting would delay value to holders; the sources describe no check on the amounts. - Upgrade risk: the 4-of-6 Safe can replace the logic immediately, and a single EOA can upgrade the rewards contract. The risk page lists this as a risk.
- MPC and operational halts: threshold compromise of the MPC nodes could cause "loss or misdirection of staked assets", and outages could pause staking and unstaking.
- Exit and peg risk: direct redemption takes a 15-day cooldown plus a 2-day window. Fast exit depends on AMM liquidity, so
sAVAXcan trade below its redemption value during stress. BENQI's own 2022 listing proposal rated market risk "B" and saidsAVAXprice is highly correlated with AVAX. (source reliability: community publication) The risk page also names redemption delays and exchange rate deviations as liquid-staking risks. - Validator risk: delegation to validators that underperform lowers rewards. The sources do not describe slashing handling or validator limits.
- Composability:
sAVAXis used as collateral in lending, so a sharp price drop or pause can trigger liquidations elsewhere. The sources do not size this exposure. - Shared keys: the 2-of-13 Safe
0x30d6…can pause bothsAVAXand BENQI lending-market functions. The Pause Guardian that holdssAVAXpause rights is itself owned by a separate 3-of-6 Safe. The measured owner lists at Avalanche block 98915447 (2026-10-30) show that all six owners of the 3-of-6 Safe are also owners of the 2-of-13 Safe. The 4-of-6 upgrade Safe0x832c…8cdhas the same six owners as the 3-of-6 Safe. Assessment: one group of six keys sits behind the upgrade, admin and resume powers, the Pause Guardian and the 2-of-13 pause Safe, so emergency control is more concentrated than the three thresholds suggest. The sources do not say who these signers are. - Effectively no deposit cap at the read block: the pooled-AVAX cap equals the maximum 256-bit integer at block 98915447 (2026-10-30), and the
setTotalPooledAvaxCaprole holder can change it.
Audit coverage for this product (separate from other BENQI products):
- Halborn, liquid staking audit, engagement 13 Nov 2021 to 14 Feb 2022: 4 critical, 1 high, 1 medium, 0 low and 1 informational finding, all marked solved. (source reliability: unverified publisher) One critical finding concerned a setter that overwrote
totalPooledAvax, and one concerned a locked contract. The scope is all contracts in the Benqi liquid-staking repository from commit 21f7e063, includingStakedAvaxand aNodeCoordinatorcontract that is not in the current contract list; the exact file list is not enumerated. Because the report was published on a host not verified as auditor-controlled, its severity counts and "solved" statuses carry lower source confidence. - Certora formal verification, 14 Mar to 14 Apr 2022, of code delivered on 23 Feb 2022. (source reliability: unverified publisher; the audit date range is unconfirmed by a stronger source)
- Aave V3
sAVAXverification and listing-steward audit: the token was formally verified against the deployed contract on Avalanche. (source reliability: unverified publisher; the claim is unconfirmed by a verified auditor publication) The audit digest names Certora as the auditor. The year of this work is not stated. The oracle adapter and listing steward in that scope were not found at a current address. - Deployed-code match: the current implementation
0xb791…fa53was not compared with any audited version, because the audited commits are not public. ThesAVAXaudit reports were published on a host not verified as auditor-controlled, which lowers confidence in both their provenance and their content. The Halborn Moonwell, Cyfrin Ignite, Dedaub Ignite, isolated-markets and Zellic oracle reports cover other products and are not counted forsAVAX. - Bug bounty: Immunefi pays up to $500,000 for critical smart contract issues, and the
sAVAXreward timelock is in scope.
Data Completeness
Based on available data, the on-chain state, mechanics and TVL are well covered, while audits of the deployed code, current yield and the off-chain operations are not.
- Available: DefiLlama TVL snapshot (2026-10-30) and history; a measured read of roles, Safes and parameters at block 98915447; BENQI architecture, risk and FAQ pages; the Certora and Halborn reports; the 2022 Aave listing proposal.
- Missing: a confirmed current APR (only an unreconciled third-party 30-day average of 5.40% and a conflicting 24.14% were found) and an exchange-rate history; how rewards are measured before
accrueRewards; slashing and validator-set rules; the MPC signer set and threshold; the staking bot's design; who the signers of each Safe are and how they are chosen (the owner addresses are measured, their identities are not stated); the rewards contract's delay; who administersgrantRoleandrevokeRole. - Limited: the
sAVAXaudits are from 2021 to 2022 or carry no stated year, and the current implementation was not compared with any audited code. The GitHub source snapshot has 823 lines, while the on-chain guard references reach about line 874, so the snapshot likely predates the deployed version (Assessment, not compared). - Stale: the 7.20% APR, the $120m DEX liquidity and the Aave parameters are from April 2022. Current DEX depth was not retrievable.
- No Blockworks (ex-Messari) reports or other articles were available for this protocol, and no hacks are recorded in the DefiLlama hacks dataset.
119 citations · click markers to view evidence
Code Security
Code Security Analysis: Benqi Staked Avax
Conclusion. Benqi Staked Avax is the sAVAX liquid staking contract on Avalanche. Its core StakedAvax code was reviewed by Halborn and twice by Certora, including a formal verification. The audits name the code they reviewed but the exact reviewed revision is not tied to the current implementation, and they carry one unresolved trust assumption. A role held by a single externally owned account (EOA, a key-controlled address) can withdraw AVAX held by the StakedAvax contract with no delay. An Immunefi bug bounty with a USD 500,000 maximum is live, and the core source is verified. Confidence is moderate: the deployed code was not compared with any audited version, and the audit PDFs sit on a host not verified as auditor-controlled.
Data Completeness
Findings ledger: 102 issue groups across 9 audit files. Only two groups are critical, high or medium and not fixed. Read in full: the ledger, the coverage map, the contract-source index, the on-chain control file, the audit digests and the audit sections and repository files quoted below. Other audit text reached this report through the ledger and digests. Sources named in the audit coverage map as left out on purpose are not counted.
- Seven of the nine audit PDFs (all except Halborn 2021 and Zellic 2025) are hosted on a Gitbook file host linked from the team's docs page; that host is not verified as auditor-controlled. This lowers confidence in claims that rest only on them, not the score. The Halborn 2021 and Zellic 2025 reports are on the auditors' own GitHub publication paths.
- Several audits state no scope or date (see the table below). The audit coverage map lists no collection gaps.
- Public commits for the audited revisions are not available, so no code comparison was made. Halborn names only the engagement's starting commit , and Certora's later report reviewed the deployed contract on Avalanche without naming a commit for it.
- TVL is about USD 200.5M on Avalanche .
Audit Coverage
Nine audit files are in the workspace. Not all concern sAVAX; the products are separated below.
Audits of the sAVAX liquid staking contract (this product)
| Auditor | Date | Scope |
|---|---|---|
| Halborn (source reliability: unverified publisher) | engagement 13 Nov 2021 to 14 Feb 2022 | StakedAvax, NodeCoordinator in the liquid-staking repository (per the digest); findings: 4 critical, 1 high, 1 medium, 1 informational, all marked fixed |
| Certora (formal verification) (source reliability: unverified publisher) | 14 Mar to 14 Apr 2022, on code delivered 23 Feb 2022 | StakedAvax, rules proved with the Certora Prover; some rules verified only under assumptions |
| Certora (verification plus listing review) (source reliability: unverified publisher) | year not stated | StakedAvax.sol formally verified against the deployed contract ; also the Aave V3 sAVAX oracle adapter and listing steward (commit 1d00da87) |
Halborn and Certora are independent firms. The Certora report says it found no major issues and that ADMIN role holders are "completely trusted" .
Audits of other Benqi products (context only, not review depth for sAVAX)
- Halborn, 3 May 2021: the Benqi lending protocol (Comptroller, Qi tokens and related contracts).
- Zellic, 8 May 2025:
BenqiDualOracleandBenqiPriceOracle. - Dedaub (28 Mar 2023) and Cyfrin (Sep 2024): Benqi Ignite validator bootstrapping.
- A review of Benqi Avalanche Ecosystem Markets (isolated lending markets), date not stated.
- Halborn, Aug 2023: titled "Moonwell Finance Contracts V2" . It names another protocol and is not counted for Benqi.
Unaudited or unconfirmed areas: the PauseGuardian scope could not be determined from the available material.
Deployed code vs audits
Deployed code was not compared with the audited versions: the protocol's own StakedAvax source that Sourcify matches at Avalanche is in scope of audits by name, but the exact reviewed revisions are not public. Share of protocol-owned verified lines: 100.0% unknown, 0.0% identical, cosmetic, changed or unaudited; 3 files and 1,331 lines are in the unknown bucket. Uncovered lines by reviewed criticality: 955 critical and 376 with unassessed criticality. Roles of the rated code: funds, access control, upgradeability, oracle and view utilities. Some code has not been assessed for criticality (PauseGuardian, 376 lines). No files lack line counts and no components are unmeasured. The map reports no collection gaps.
| Component | Deployment | Status | Audit | Criticality | Note |
|---|---|---|---|---|---|
PauseGuardian | Avalanche 0xaad2d03809430eaf2ec17bc6f4d2260e6e409cb7, Sourcify runtime match | the audit scope could not be determined from the available material | none confirmed | unrated (376 lines unknown) | Reads the lending Comptroller; owned by a Safe (threshold 2 of 13 signers on-chain) |
StakedAvax (sAVAX) | proxy 0x2b2c81e08f1af8835a78bb2a90ae924ace0ea4be, implementation 0xb791c7a42fd0d10f90deaa906a8735f79719fa53, Sourcify runtime match | named in audit scope, but the audited version cannot be matched to the deployed one | Certora (twice), Halborn | critical: funds, access control, upgradeability, oracle, view utilities (955 lines) | version not compared: the exact audited revision is not public |
| 39 named contracts (Comptroller, QiToken family, rate models, Timelock, GovernorAlpha, sAVAXOracleAdapter, AaveV3SAVAXListingSteward and others) | no attributed address in the coverage map | named in audit scopes; the coverage map found no current deployment address | Halborn 2021 and Certora | not rated | mostly Benqi lending contracts, another product; inventory is partial |
The address inventory is partial: the coverage map notes. The on-chain control file separately attributes Comptroller to 0x486af39519b4dc9a7fccd318217352830e8ad9b4 on Avalanche through the PauseGuardian getter, a lending contract of another product; that address is not matched to the audit scope in the coverage map. A TransparentUpgradeableProxy at the sAVAX address is also matched by Sourcify at runtime and creation. Neither a source match nor audit scope equality proves the absence of vulnerabilities.
Findings by Severity
Critical
Halborn's four critical findings on the liquid-staking contracts (invalid setter causing pool inconsistency, memory instead of storage, contract locked, contract missing owner) are all marked fixed in the ledger and are trusted as fixed.
High
- Admin role fully trusted with unrestricted deposit and withdraw (Certora, 2022) — — Audit status: not stated; severity is High (estimated): the audit states no rating, and this AI estimate rests on the audit's statement that admins can withdraw AVAX from the pool (an assessment of the audited code; the current share of pooled AVAX held in the contract was not measured). — Check: the latest public repository code (Benqi-fi/BENQI-Smart-Contracts@e0cfd24) may still contain the code this finding describes; confidence is moderate because the role holder is outside the files, and deployed code was not compared. Deployed-state context: on Avalanche at block 98915447 (the on-chain read is partial, with 2 privilege entries unresolved) the
ROLE_WITHDRAWholder is0x565f0fe9715e3cb0df579f186c299d6707887e83, an address with no code (a single key), and the withdraw function has no delay . This is a trust assumption the team has not removed; users depend on that key's honesty. Assessment: this is the largest code-level trust risk in the product. - Other high findings (Halborn 2022 out-of-bounds access; Dedaub 2023 and Cyfrin 2024 Ignite findings; Halborn Moonwell routers, not counted for Benqi) are marked fixed.
Medium
- Missing Payment Amount Check (Halborn 2021, Benqi lending, another product) — — Audit status: acknowledged — Check: the code was rewritten; the check could not determine whether this finding still applies.
- All other medium groups (Halborn 2022 timestamp underflow, Dedaub, Cyfrin, Zellic staleness check, isolated-markets review) are marked fixed: 15 groups.
Low / Informational
The ledger holds many low and informational groups, mostly fixed or acknowledged gas, style and validation points. Low groups still acknowledged include the Halborn 2021 block.timestamp and inline-assembly notes, the Cyfrin Chainlink feed validation and slippage notes, and the isolated-markets precision and uncapped seize-share notes. Informational groups without a stated status include the Certora gas suggestions and the Zellic transferOracleAdmins note.
Recurring Patterns
- Trusted-admin or privileged-role risk recurs: the Certora ADMIN trust assumption (22583), the Cyfrin finding that the default admin controls all roles (fixed), and the Cyfrin note on misaligned pause controls (acknowledged).
- Oracle validation recurs across products: Dedaub (2023), Cyfrin (2024) and Zellic (2025) each reported staleness or feed-validation gaps; the high and medium ones are marked fixed, while Cyfrin's low-severity Chainlink validation finding is acknowledged.
- No single finding group is reported by several audits.
Code Quality Assessment
- Review depth for sAVAX is good in kind: two firms, plus formal verification with named properties. The reports are older than the current period and the link to the deployed version is not shown, so the deployed code is not shown to match.
- All critical findings and all high findings on sAVAX are fixed except the Certora ADMIN-trust assumption. That open item is architectural: a key-held role can withdraw AVAX held by the
StakedAvaxcontract. The sAVAX proxy admin is a Safe with threshold 4 of 6 and zero fastest delay on upgrades . A second contract that holds theROLE_ACCRUE_REWARDSrole,0x09c1e991870cbc01009a4b49397a4f2a127d3784, is an upgradeable proxy whose upgrade rights sit with an EOA (0x0df1a01ade3cd67ccc11d89f2859a0de514cd679), also with zero fastest delay. - Assurance: Immunefi lists a maximum bounty of USD 500,000, live since 19 August 2021 and updated 05 August 2026 . The core source is verified by Sourcify. The bounty is below 1% of TVL. Its default out-of-scope list excludes "Impacts involving centralization risks" and impacts requiring access to privileged addresses , so it does not cover the ADMIN-trust finding above.
- Gaps: no public audited revisions, undetermined
PauseGuardianscope, and 39 audit-named contracts without an address in the coverage map.
Score: 6.0/10
Rubric security-score-v2.6; data completeness 1.00
- Audit coverage
- 5.50 points
- Audit depth
- 7.50 points
- Unresolved findings
- 9.88 points
- Assurance
- 7.50 points
- High · not stated
- 1 findings
- Medium · acknowledged
- 1 findings
| Factor | Level | Evidence | Weight | Points |
|---|---|---|---|---|
| Audit coverage (computed) | L2 | supported | 0.35 | 5.50 |
| Audit depth | L3 | supported | 0.20 | 7.50 |
| Unresolved findings (computed) | L4 | supported | 0.30 | 9.88 |
| Assurance | L3 | supported | 0.15 | 7.50 |
19 citations · click markers to view evidence
Economic Security
Economic Security Analysis: Benqi Staked Avax
Conclusion. Benqi Staked Avax is the sAVAX liquid staking token on Avalanche. Its documented yield is AVAX staking rewards; the current share of any token emissions in returns is not documented. Its main economic weakness is custody and control. User AVAX leaves the C-Chain contract to be staked through an off-chain MPC (multi-party computation) signing set. Admin roles can withdraw pooled AVAX and pause redemptions with no stated limit. Nothing in the workspace shows an insurance fund, a slashing rule, or an independent proof of reserves. Confidence is moderate: the contract mechanics and on-chain roles are well evidenced, but exit liquidity and holder concentration are only weakly sourced.
Scope: the DefiLlama record covers one product, the sAVAX staking contract (0x2b2C81e08f1Af8835a78Bb2A90AE924ACE0eA4bE) on Avalanche. Benqi Markets (lending), Ignite (validator service) and the QI token are other BENQI products. They are context only unless a shared component is named.
TVL & Liquidity Profile
TVL is about USD 200M, all on Avalanche, and the recent decline is mostly a price effect.
- The DefiLlama snapshot shows , collected 2026-10-30 . The only chain is Avalanche , so its share is 100%.
- Chain profile: Avalanche is classed
kind: otherwith no L2BEAT stage and no exit window . It is a base chain, not a rollup. No chain launch date is cited. - 30-day change is USD -35.8M, of which USD -41.5M is price effect . Token quantity rose by about USD 5.7M at current prices, so the 30-day fall reflects the AVAX price, not net redemptions.
- Over 90 days TVL rose USD 70.7M, of which USD 57.3M is price effect .
- On-chain at block 98915447 (Avalanche C-Chain, 2026-10-30), the contract reports
stakerCount: 64761. The exchange rate from the reportedtotalPooledAvaxandtotalSharesis about 1.29 AVAX per sAVAX. - Exit path: the protocol redemption has a 15-day cooldown followed by a 2-day claim window. On-chain values at the same block are and . The docs describe the same exit and an instant AMM swap alternative .
- DEX depth:
aggregators/dex_depth.yamlhas no measured assets (Ethereum only; sAVAX has no Ethereum listing in the workspace). Depth at 2% and at 0.5% price impact is therefore a gap, not zero. - Pool figures from GeckoTerminal (source reliability: unverified publisher) show two sAVAX/WAVAX pools with and . These are single-pool figures, not a full venue survey, and they cannot set a rating.
- Holder concentration (Snowtrace, source reliability: unverified publisher): the top two holders are contracts. One is labelled "BENQI: Benqi sAVAX" at . The other is labelled "Aave Avalanche SAVAX" at . They look like lending-market receipt contracts holding user deposits, not single whales. The first ten listed rows sum to about 87%, so most sAVAX sits in downstream lending markets. This is unconfirmed by a non-weak source.
Oracle Risk
sAVAX has no external price oracle for its own mint and redeem. Its exchange rate is internal accounting pushed up by role holders, and the report cannot cite a bound on each update in the deployed version.
- Version note: the published
StakedAvax.soldocs copy is an older GitHub master and differs from the deployed contract. The docs copy hasaccrueRewards(uint amount), while the on-chain control read shows the deployed entry asaccrueRewards()with no argument . The docs copy is therefore not used here as evidence of how the deployed contract updates the rate. - In the older docs copy the rate is
totalPooledAvax / totalShares. The deployed update rule is not established from citable text. - In the deployed version, the on-chain read shows
accrueRewards()is the rewards entry and is gated byROLE_ACCRUE_REWARDS. No citable text for the deployed version states whether one call is bounded in size, so the update-size limit is not disclosed. - The holders of
ROLE_ACCRUE_REWARDSare . They are a Safe multisig and a "sAVAX reward timelock" contract at0x09c1e991870cbc01009a4b49397a4f2a127d3784. The contract's proxy admin chain ends in an EOA (a single-key account) that can upgrade withfastest_delay_seconds: 0. - The 2022 Certora review of the Aave V3 price adapter found the rate is computed by "tracking the amount of deposited Avax tokens, as opposed to the balance of Avax held by the contract" . For the adapter version that was audited, this closes the donation-inflation path of the xSushi type. The audit host is unverified (see Data Completeness).
- The rate has no on-chain link to P-Chain validator balances, and the P-Chain stake cannot be verified from the C-Chain contract. The workspace documents no slashing handling.
- Downstream price feeds (Chainlink sAVAX/USD, the Chaos Labs dual oracle used by Benqi Markets) price sAVAX for lending and belong to those products. The Zellic oracle review of May 2025 found a missing staleness check, status fixed . That is context for Benqi Markets, not for the sAVAX contract.
- No oracle-caused loss is recorded for sAVAX in the workspace.
Flash Loan & MEV Exposure
No flash-loan or MEV vector against the sAVAX contract is documented; the main exposure is at the AMM exit.
- Minting and redemption use the internal rate, not a DEX spot price. The rate is changed only by holders of the rewards role, not by an arbitrary caller.
- Redemptions are time-locked. The older docs copy (not confirmed to match the deployed version) shows
_redeempaying at the exchange rate stored closest to the unlock time . The 15-day cooldown read on-chain (see above) is what prevents a same-block mint-and-redeem; the docs-copy function is context only. - Users who skip the cooldown trade sAVAX on AMMs and face slippage and sandwich risk on thin pools. Pool depth is only weakly sourced (see above).
- The Halborn liquid-staking review (Nov 2021 - Feb 2022) lists a critical, fixed bug where
setTotalPooledAvaxCapsettotalPooledAvaxinstead of the cap . The audit text is tagged as fixed, but the host is unverified. - Liquidation front-running belongs to Benqi Markets and Aave, not to the sAVAX contract.
Tokenomics Risk: Yield, Leverage & Reflexivity
The documented sAVAX yield is AVAX staking rewards with a 10% protocol cut, and the sAVAX contract has no borrowing; the current QI emission share is not documented.
- Yield source: staking rewards on the P-Chain accrue to the sAVAX rate .
- Cited amounts: DefiLlama fees are USD 1.03M over 30 days and USD 12.6M over 365 days; revenue (the protocol share) is USD 102,955 and USD 1.26M .
- The on-chain
protocolRewardShareis 0.1 (10%) . This matches the revenue-to-fees ratio. Users keep about 90% of rewards. - Team target (dated): a 2022 Benqi post on the Aave forum (source reliability: community publication) says validation generates "approx 7.20% APR" . It also says "No emission schedule" for sAVAX . Both are a team statement from April 2022, not current data, so the current emission share is unknown. The current BENQI docs describe sAVAX as a token that "accrues interest from Avalanche staking rewards" and name no QI share of that yield. A web check found no current QI reward speed for sAVAX pools or Benqi Markets and no dated QI emission schedule, because the lending app pages failed to render. The QI emission share of downstream yield therefore stays an open gap.
- Own token in backing: sAVAX is backed by pooled AVAX. QI is not part of the backing, and minting or redeeming sAVAX does not touch QI. QI is 7.2B supply with the community allocation "fully unlocked" . It does not back sAVAX, so unlocks are not scored here.
- Leverage: the sAVAX contract has no borrowing. Users who loop sAVAX in Benqi Markets or Aave take on leverage in those products, with the collateral factors and liquidation rules of those markets (context only). A pool cap
totalPooledAvaxCapexists. It is currently the maximum uint, so it is effectively unbounded . - Reflexive loss path: sAVAX trades at a discount if the exit queue is slow or paused. Lending-market contracts that hold a large share of the supply (unverified-publisher holder data) could then liquidate sAVAX collateral into thin pools. This is a plausible scenario from the structure, not an observed event.
Solvency, Liquidation & Bad Debt
Backing is pooled AVAX that is sent off-chain for staking, with a 15-day exit, an unbounded admin withdraw and pause, and no cited loss cover.
- Backing: the contract tracks
totalPooledAvaxin the on-chain views at block 98915447. Docs say pooled AVAX "is then delegated to validator nodes on the Avalanche P-Chain" . The delegated share is not quantified. Docs also say the staking AVAX is bridged "via MPC encryption" . No independent attestation of the P-Chain stake or of reserves is in the workspace. - Admin withdraw:
ROLE_WITHDRAWcan move AVAX out of the contract for delegation, with no cap in the code. On-chain, the only holder is an EOA,0x565f0fe9715e3cb0df579f186c299d6707887e83, with no delay . The 2022 Certora formal-verification report states that admins can "deposit and withdraw AVAX from the system without limitation" . A 2022 Benqi post (source reliability: community publication) says the delegating EOA ownership "is in the process of being transitioned to a MPC" model . Whether the currentROLE_WITHDRAWEOA is MPC-controlled is not documented. - Exit bound: the 15-day cooldown and 2-day window are not a hard bound. On-chain,
setCooldownPeriod(uint256)andsetRedeemPeriod(uint256)are callable by the 4-of-6 admin Safe withfastest_delay_seconds: 0. Assessment: the stated exit time can be lengthened by the admins without notice, which supports the low rating of the exit path. - Pause: docs list the ability to "pause staking or redemptions" . In code,
_redeemiswhenNotPaused. On-chain, pause roles sit with a 2-of-13 Safe and a 4-of-6 Safe, with delay 0. No maximum pause duration is stated.pausedandmintingPausedwere both false at block 98915447. - MPC halt: if too many MPC nodes are down, "staking, unstaking, and cross-chain transfers will be temporarily paused" . No time bound is given.
- Loss handling: no insurance fund, reserve, or slashing rule is documented for sAVAX. A web check for an sAVAX insurance or compensation fund found nothing; a BENQI Safety Module is mentioned only in a summary with no contract, size or rules, so it is not treated as a fund.
aggregators/defillama_hacks.yamllists no incident records, so there are no payouts. - Liquidation and bad debt: the sAVAX contract holds no debt positions. Liquidation and bad-debt records for Benqi Markets and Aave belong to those products and are not assessed here.
Incentive Alignment
The protocol earns 10% of staking rewards, so its revenue grows with TVL; the weak point is that depositors cannot independently check custody.
- The team earns from the same reward flow users receive (10% share), which aligns it with TVL growth. The protocol-share recipient on-chain is
0x3d80e0a69cf26cdbccf90131912f3f85106613cf. - Mercenary-capital risk cannot be sized: the documented sAVAX yield is staking rewards, but current QI incentives on sAVAX pools and in Benqi Markets were not found.
- No death spiral by design: sAVAX is a claim on AVAX, not an algorithmic token. The tail risk is a combined AVAX price fall and exit freeze, as described under Tokenomics.
- Loss cover: no fund is evidenced (funds list is empty). Size, trigger and payout rules are unknown. The Immunefi bug bounty (maximum USD 500,000) is not a loss-cover fund .
- Past payouts: no incident records exist for this protocol family in the DefiLlama hacks file, so no past payout is evidenced.
Economic Attack Vectors
The credible attack paths are custody-layer and admin-layer, not price manipulation.
- Key compromise at the withdraw role: the role holder is a single EOA with no delay. A compromise could move pooled AVAX out of the contract. Impact: up to the liquid AVAX balance in the contract at that time. Assessment: likelihood is low to moderate, because the key's custody setup is not documented. Evidence that the EOA is an MPC-controlled account would move it to low.
- MPC threshold breach: docs state that an attacker who breaches the threshold "could generate valid signatures to cause loss or misdirection of staked assets" . The share of pooled AVAX exposed to this is not quantified, and no cap is cited.
- Upgrade risk: the sAVAX proxy is upgradeable by a 4-of-6 Safe with
fastest_delay_seconds: 0. Docs state that "Certain protocol contracts, including those governing the lending markets and liquid staking, are upgradeable." - Rate manipulation: only role holders can update the rate (see Oracle Risk). No source documents how a slashing loss on the P-Chain would be handled.
- Liquidation cascade through downstream markets: most sAVAX supply sits in lending-market contracts (weak source). A sharp AVAX or sAVAX discount could force liquidations into pools of USD 0.1-0.5M liquidity (weak source). This is a structural risk; no event is recorded.
- Bridge risk: the only cross-chain movement is the C-Chain to P-Chain AVAX transfer inside the Avalanche network, run by the MPC set. No third-party bridge is cited.
- Critical dependencies for the dependency view:
- the MPC signing set that moves AVAX between C-Chain and P-Chain, with no cited cap;
- the validators that receive the delegation (not enumerated in the workspace);
- the Safe multisigs and the EOA that hold the roles;
- Avalanche itself, the base chain.
- The dependency graph file lists no protocol dependencies: one
LIVES_ONedge to Avalanche at share 1.0, empty composition (status: missing) and no warnings .
Aggregator Scores
No curated security-score files are in the workspace.
aggregators/certik.yaml,aggregators/defisafety.yamlandaggregators/exponential.yamlare absent, so no CertiK, DeFiSafety or Exponential assessment is compared.- The DefiLlama TVL is the latest captured snapshot (
collected_at2026-10-30), as given above. aggregators/onchain_control.yamlis code-measured state for Avalanche C-Chain at block 98915447 (2026-10-30). Its role holders, thresholds and cooldown values are used above as current values.
Data Completeness
Mechanism, roles and fees are well evidenced; reserves, liquidity depth, holder concentration and loss cover are not.
Available:
- DefiLlama TVL, history and fees.
- On-chain role and parameter reads.
- The sAVAX source (docs copy). It is an older version than the deployed contract (the rewards function signature differs), so it is context only.
- Official risk docs.
- Nine audit files. Seven are hosted on unverified sites (
unknown_host_unverified) per the audit-trust warnings. The Zellic (May 2025) and Halborn (2021) reports are marked as trusted auditor sources in the audit index. Claims resting only on the seven unverified audits carry lower confidence; their provenance is not treated as a protocol penalty.
Missing or weak:
- Independent proof of the P-Chain stake and MPC key arrangement.
- DEX depth at 2% and 0.5% impact:
aggregators/dex_depth.yamlhas no assets. - Non-weak holder-concentration data: holder figures come from an unverified-publisher page.
- A slashing or loss-handling rule.
- Any insurance fund: the web check found none.
- Current QI incentives on sAVAX pools: a web check found no reward speed or emission schedule, and the lending app pages did not render.
- Current dependency scores: none, because the graph lists no protocol dependencies.
Confidence in the assessment: moderate (about 0.6).
Score: 3.8/10
Rubric security-score-v2.6; data completeness 0.65
- Solvency redemption
- 3.50 points
- Oracle design
- 3.50 points
- Dependency risk
- 3.50 points
- Liquidity concentration
- 3.50 points
- Yield design
- 5.50 points
| Factor | Level | Evidence | Weight | Points |
|---|---|---|---|---|
| Solvency redemption | L1 | supported | 0.35 | 3.50 |
| Oracle design | — | not disclosed | 0.25 | 3.50 |
| Dependency risk | L1 | supported | 0.15 | 3.50 |
| Liquidity concentration | — | not disclosed | 0.10 | 3.50 |
| Yield design | L2 | supported | 0.15 | 5.50 |
45 citations · click markers to view evidence
Team & Transparency
Social & Team Transparency: Benqi Staked Avax
Benqi Staked Avax (sAVAX) is the liquid staking product of the BENQI protocol on Avalanche. The official documentation names no individuals behind the team. The official benqi.fi site links a Terms of Service document that names an operating company with Panama as governing law, but the document is hosted on Google Docs (source reliability: unverified publisher), so those details stay unconfirmed. Founder names come only from third-party or self-published pages (source reliability: unverified publisher or community publication), and those lists conflict. Disclosure is good: contract addresses, audit reports and a bug bounty are public. Development is low-volume but alive, and no hack is recorded for this product. No named person or entity is verified by an official or established source, so confidence in the identity assessment is moderate.
Team Identity
Official BENQI documents name no team member. Only weak sources name people or an operating company, and their lists conflict.
-
The official docs say only that the founding team governs the protocol:
-
The token-distribution page refers to "team members" without naming them:
-
Aave governance posts come from the account "Team_BENQI". The author is described only as "a representative of the BENQI team" (source reliability: community publication):
-
The Rome Blockchain Labs (RBL) site (source reliability: unverified publisher) says the founders launched BENQI in August 2021:
-
The same page lists Hannu Kuusi, JD Gagnon and Eino Mäkitalo as co-founders and Alexander Elhorst as "Executive". It also names three people who met as independent consultants at Dextools:
-
That timeline names "Alexander Szul" while the team list names "Alexander Elhorst". The source does not explain the difference, and the RBL page is undated.
-
A 2021 press reprint (source reliability: unverified publisher) names the founders as JD Gagnon, Hannu Kuusi and Alexander Szul:
-
An Uphold blog post (source reliability: community publication) says BENQI "was founded by Rome Blockchain Labs Inc." and names Dan Mgbor (operations), Hansen Niu (strategy) and Jason Tuang (DeFi): iq.wiki (source reliability: unverified publisher) lists JD Gagnon, Dan Mgbor and Hannu Kuusi. These sets differ and none is confirmed by an official or established source.
-
The iq.wiki page (source reliability: unverified publisher) says BENQI "is developed in collaboration with Rome" Blockchain Labs. No official BENQI document confirms this link.
-
The official benqi.fi site footer links its Terms of Service:
-
Those terms (updated April 4, 2026; source reliability: unverified publisher, a Google Docs copy) name the operator, which stays unconfirmed: They name the Decentralized Technology Foundation as its sole shareholder. The terms are governed by Panama law with exclusive jurisdiction of courts in Panama: The place of incorporation is not stated, and no registry filing was found. No natural person is named as an officer or owner. Because the publisher is unverified, this document does not set the identity level.
-
The 2022 Aave proposal named non-team multisig signers: Mark Stanwyck (Avalaunch), "TN" (Pendle) and Marc Zeller (Aave) (source reliability: community publication): This describes 2022 signers. Current signers are not confirmed here.
Assessment: the team has operated under the BENQI brand since August 2021, so it is not a fresh anonymous group. Entity and individual names rest on weak sources only, so no named leader or legal entity is verified.
Track Record
The hacks dataset records no hack for this product or its family; one May 2026 security event in a sibling lending market appears only in weak sources.
- The hacks dataset returned no records: The incidents list is therefore empty.
- The TVL history lists the product from 2022-02-17 with TVL of USD 57.6M:
- TVL peaked in the series near USD 528M in October 2025 and was about USD 200.5M on the latest snapshot:
- An iq.wiki article (source reliability: unverified publisher) says Chaos Labs detected suspicious activity on its operational wallets on May 4, 2026, and BENQI paused its Avalanche Ecosystem Markets: The article says BENQI announced on May 14, 2026 that it would wind down those markets except NXPC, that Chaos Labs reported "no user funds were lost", and that Core Markets moved to Chainlink-only feeds:
- A focused web search found a BENQI statement on X (source reliability: unverified publisher). It says BENQI was notified by Chaos Labs on May 4 and "Fully paused the Ecosystem Markets powered exclusively by Chaos Edge Oracle": The same statement says: It also says: These are the project's own claims, not independently verified.
- A Chaos Labs statement on X (source reliability: unverified publisher) says:
- The event concerns the Ecosystem Markets, a sibling lending product, and the captured statements do not name the sAVAX contract as affected. The captured statements do not address bad debt. No formal post-mortem was found, the wind-down is reported only by iq.wiki and remains unconfirmed, and the hacks dataset has no record. There is no source above the weak tier, so the event is not listed as an incident.
- Review work specific to sAVAX is a Halborn liquid-staking audit and a Certora formal verification of the
StakedAvaxcontract. The Halborn report lists its engagement: The Certora report says the work ran "from March 14, 2022 to April 14, 2022": Both reports are hosted by the project, and their publication is not verified as auditor-controlled: - RBL's timeline lists earlier projects by the founders: Velox (January 2021), Opulentia (July 2021) and Kinetic on Flare. These are context only (source reliability: unverified publisher).
- Own token depeg history:
aggregators/asset_checks.yamlfound no measurable Ethereum tokens for this protocol, because sAVAX lives on Avalanche. No depeg measurement exists, so no token rows are listed. - The available records do not establish a history of pauses, outages, exit halts or depegs for sAVAX, in either direction. A pause role exists on the contract, and the on-chain read shows it unpaused today:
Social Presence
GitHub shows low but current activity. Other channels have only partial data.
- The main smart-contracts repository's last visible commit is January 2023 ("Add PauseGuardian contract"), which is older than 12 months.
- The
avalanche-monitoringrepository in the same GitHub organization has commits on 2026-02-05, 2026-02-09, 2026-02-11 and 2026-02-13, then December 2025, September 2025 and July 2025. Commits fall in 2 of the last 12 months. - The captured X page of the BENQI account (source reliability: unverified publisher) shows a post dated 2026-10-23 with 2,410 views: The same page shows the account's follower count: The count was captured when the page was fetched (2026-10-29).
- In 2022 the team reported 62.9k Twitter followers and 13.35k Telegram followers (source reliability: community publication): These numbers are old. No current Discord or Telegram member counts were collected.
- Docs link a Discord invite and a Twitter account for updates:
- On-chain, the sAVAX contract reports 64,761 stakers in its
stakerCountview: - The sources include no blog or Medium archive, so post frequency cannot be judged.
Transparency Indicators
Contracts, audits and the bounty are published. No treasury or revenue reporting was found.
- Addresses: the docs list the sAVAX address: The docs page is marked partial, and the list is not shown to be complete.
- Audits: the docs list Halborn (May 2021, and November 2021 to February 2022), Certora (June 2022), Dedaub (March 2023), Cyfrin (December 2024), a 2024 Isolated Markets audit and a Zellic oracle audit (May 2025): Several of these cover other BENQI products (Ignite, markets, oracle), not sAVAX. One linked Halborn report is titled for Moonwell Finance, a family product.
- Bug bounty: an Immunefi program is live since 19 August 2021, last updated 05 August 2026, with maximum bounty of USD 500,000: The critical payout is "capped at 10% of the economic damage funds at risk" with a floor of USD 50,000. A sAVAX reward timelock address is in scope:
- Open source: the contracts repository is public: Sourcify reports a runtime match for the sAVAX proxy:
- Risk partners: the docs say BENQI works with Chaos Labs on risk management:
- Financials: treasury, expense and revenue reports from the team are not in the available sources. Aggregator data gives protocol fees of USD 1.03M over 30 days (code-collected, not a team report):
- Incident reporting: no post-mortem exists in the sources, and no confirmed incident requires one.
Red Flags
- No named team member in official documents; founder and leader names in weak sources conflict, and the operating company named in the Terms of Service has no stated place of incorporation.
- A name discrepancy (Szul versus Elhorst) on the RBL page.
- The main contracts repository has no commit since January 2023; recent commits are in a monitoring tool.
- May 2026 oracle-provider wallet compromise and pause of a sibling lending market, with a wind-down reported by iq.wiki. Only weak sources (source reliability: unverified publisher) cover it, no formal post-mortem was found, and bad debt is not addressed. The project's own statement says no user funds or positions were impacted.
- No periodic treasury or reserve report found.
Positive Signals
- Live product since 2022-02-17 with no record in the hacks dataset.
- A Halborn liquid-staking audit and a Certora formal verification of the
StakedAvaxcontract (both project-hosted reports, publication not verified as auditor-controlled), and a USD 500,000 Immunefi bounty maintained through 2026. - First-party contract address list and public, Sourcify-matched source for the sAVAX proxy.
- The team publicly proposed expanding the 2022 multisig to include outside signers (source reliability: community publication).
- GitHub commits in 2026, and an X post on 2026-10-23 seen on a captured page (source reliability: unverified publisher).
Data Completeness
- Available: official docs, the benqi.fi site, Immunefi bounty page, GitHub commit pages, X pages, the Terms of Service copy, the RBL website, iq.wiki, press reprints and Aave forum posts.
- Missing: named leaders from an official or established source, the operating company's place of incorporation or any registry filing, a formal post-mortem for the May 2026 event, current Discord and Telegram sizes, blog history, treasury reports, post-mortems.
- Two focused web searches were run: one for current leadership and the legal entity, which found the Terms of Service, the "founding team" statement and the weak founder lists above but no dated official leadership list; and one for the May 2026 event, which found the project and Chaos Labs statements on X but no formal post-mortem. Two press pages on the Chaos Labs wallet attack do not mention BENQI. Social data comes from collected pages, not live scraping.
Score: 4.0/10
Rubric security-score-v2.6; data completeness 1.00
- Team identity
- 1.00 points
- Disclosure
- 7.50 points
- Development activity
- 5.50 points
| Factor | Level | Evidence | Weight | Points |
|---|---|---|---|---|
| Team identity | L0 | supported | 0.45 | 1.00 |
| Disclosure | L3 | supported | 0.35 | 7.50 |
| Development activity | L2 | supported | 0.20 | 5.50 |
Track record: 8.6/10
Rubric security-score-v2.6; data completeness 0.80
- Operating history
- 9.50 points
- Incident record
- 10.00 points
- Resilience
- 3.50 points
| Factor | Level | Evidence | Weight | Points |
|---|---|---|---|---|
| Operating history (computed) | L4 | supported | 0.30 | 9.50 |
| Incident record (computed) | L4 | supported | 0.50 | 10.00 |
| Resilience | — | not disclosed | 0.20 | 3.50 |
41 citations · click markers to view evidence
Governance
Governance Analysis: Benqi Staked Avax
Conclusion. BENQI's liquid staking token sAVAX on Avalanche C-Chain is run through multisig wallets and single keys, with no timelock found. BENQI documents Miles voting only for validator delegation, and no vote on upgrades or parameters was found. The upgrade path is a 4-of-6 Safe multisig whose signer identities no source confirms. Separately, a single code-less address (EOA, or possibly an off-chain multi-party-computation wallet) holds the role that withdraws AVAX from the contract. Users depend on the operational security of those signers and key holders. Evidence is strong for the on-chain control state, and weaker for who the signers are and how the withdraw key is secured.
Scope: this report covers BENQI Liquid Staking (sAVAX, StakedAvax proxy 0x2b2c…a4be) on Avalanche C-Chain. BENQI's lending markets share some operators (the 2-of-13 pause Safe), but their own contracts are not assessed here.
Governance Structure
BENQI states that protocol parameters and liquid staking functions are updated through a multisig framework, and that the founding team currently governs the protocol.
- Type: multisig-run; BENQI's documentation says "The founding team currently governs the protocol." BENQI says it "distributes signing authority across multiple parties through a multisig framework" and that "liquid staking related functions" may be updated through it .
- Voting: BENQI Miles holders do vote, but on validator delegation. Node Voting lets them "vote regarding how AVAX is delegated from BENQI’s Liquid Staking Pool" . It covers the BENQI Miles Pool (35% of delegations) only . Voting on interest rate models and risk parameters is listed only as a possible future use, "subject to change" . No source found establishes a vote on
StakedAvaxupgrades, roles or fee parameters. - Vote concentration: no voting venue (Snapshot space or Governor) was measured; the measured venue list is empty . No quorum or top-voter figures exist, and a web search found no BENQI Snapshot space or Governor/Timelock address for sAVAX control.
- Timelock: none on the sAVAX upgrade or role functions. The on-chain read gives
fastest_delay_seconds: 0for upgrade, pause and fee functions (see Admin Key Analysis). - Emergency bypass: a separate 2-of-13 Safe and a guardian contract can pause, and the 4-of-6 upgrade Safe also holds the pause role; all act instantly (below).
Admin Key Analysis
Several keys can act instantly. The upgrade and admin powers sit with a 4-of-6 Safe, pause powers with a 2-of-13 Safe, and the AVAX-withdraw role with one code-less address. On-chain state is for Avalanche C-Chain at block 98915447 (2026-10-30).
| Power | Holder (type, threshold) | Delay |
|---|---|---|
Upgrade StakedAvax proxy (via ProxyAdmin 0x2295…5a66) | Safe 0x832c…e489cd, 4-of-6 | 0 |
DEFAULT_ADMIN_ROLE: set protocol reward share and recipient, resumeMinting, accrueRewards; likely grant/revoke of every role (see below) | same 4-of-6 Safe | 0 for the listed functions; grant/revoke delay not measured |
pause() | 4-of-6 Safe, 2-of-13 Safe 0x30d6…9aa99159, contract 0x10f3…82e6 (owned by a 3-of-6 Safe 0xb952…db3d) | 0 |
pauseMinting() | 2-of-13 Safe, contract 0x10f3…82e6 | 0 |
withdraw(uint256) (sends contract AVAX to the caller) | one address with no code, 0x565f…7e83 | 0 |
Upgrade reward-accrual contract 0x09c1…3784 | owner of its ProxyAdmin is an address with no code, 0x0df1…d679 | 0 |
- Upgrade Safe: the proxy admin is
ProxyAdmin0x2295…, whose owner is the 4-of-6 Safe . The Safe has no modules and no guard, so M-of-N is the only path. Signer addresses are listed on-chain, but no source names the people behind them. A web search for current signers returned no roster; the Safe interface did not load owner data. - Pause Safe: a 2-of-13 Safe can pause and pause minting . It also owns the
PauseGuardianfor lending markets. The pause roles are limited to pausing, which is a reasonable emergency design, but two signatures suffice. - Withdraw role:
ROLE_WITHDRAWforwithdraw(uint256)is held by an address withcode: false. This is single-key control as far as on-chain data shows. BENQI says it moves AVAX to the P-Chain through an off-chain MPC (multi-party computation) system . A 2022 forum post says the "delegating EOA ownership is in the process of being transitioned to a MPC (multi-party computation) model" (source reliability: community publication) . Assessment: this address is probably the MPC-controlled operator wallet, but its threshold is not disclosed. - Reward contract:
0x09c1…3784holds theROLE_ACCRUE_REWARDSrole onStakedAvaxand is listed in BENQI's Immunefi scope as "sAVAX reward timelock" . Its proxy admin is owned by a code-less address, and the upgrade delay is 0 . - Docs vs chain: a forum post says the sAVAX token "is currently held in a 4/7 multisig by the team" and that the multisig would expand to 10 signers including Aave, Pendle and Avalaunch representatives (source reliability: community publication) . The statement is from a 3 May 2022 post by the BENQI team in an Aave governance thread, so it is historical. On-chain at block 98915447 the upgrade Safe is 4-of-6, not 4-of-7 or 10 signers. A web check found only this 2022 plan and no current roster, so whether any current signer is outside the team is not confirmed.
- Risk disclosure: BENQI itself warns that compromised signers could "Disable or interfere with protective measures" and redirect staked AVAX .
- Benefit: the 2-of-13 Safe gives a lower-threshold, wider pause path that can respond to an incident without holding upgrade power. The pause role is not fully separated from upgrades, because the 4-of-6 upgrade Safe also holds it.
Upgradability
StakedAvax is an upgradable EIP-1967 proxy that its admin can change instantly.
- Pattern:
StakedAvax0x2b2c…a4beis anEIP1967Proxywith implementation0xb791…fa53. The sAVAX token logic and share accounting live in this proxy, so an upgrade can change balance logic. - Who and how fast: the 4-of-6 Safe via
ProxyAdmin, no timelock (fastest_delay_seconds: 0). - Exit window: unstaking has a cooldown of 1296000 seconds (15 days) and a redeem period of 172800 seconds (2 days) . BENQI also documents a "15-day cooldown period" . An upgrade with zero delay lands well inside a 15-day exit, so users cannot exit before it takes effect (secondary market sales aside).
- Verification: the proxy matches Sourcify.
Parameter Controls
Admin roles can change fee, cap and cooldown parameters with no cited bounds or delay.
- The protocol reward share is currently
protocolRewardShare: '100000000000000000'(10% if 1e18 is 100%), and the admin Safe can change it and its recipient . The recipient is a 3-of-9 Safe. - The pooled-AVAX cap is currently set to the maximum uint256 (effectively no cap);
ROLE_SET_TOTAL_POOLED_AVAX_CAPis held by the 4-of-6 Safe. - Source-code reads show
setCooldownPeriodis gated byDEFAULT_ADMIN_ROLEwith no visible bounds check or timelock (inference from code in the digests) . Assessment: the admin Safe could lengthen the exit cooldown, which would hurt users without a way to leave first. withdraw()sends contract AVAX to the caller after only a role check, with no amount limit visible .- Role administration: the contract source sets
DEFAULT_ADMIN_ROLEininitializeand inherits OpenZeppelinAccessControlUpgradeable. Assessment: the default admin, the 4-of-6 Safe, can therefore grant and revoke every role includingROLE_WITHDRAW, since no custom role admin appears in the source. The on-chain read did not resolve the per-role admin, and the delay for grant/revoke is not measured.
Decentralization Assessment
Control sits with a small set of multisig signers and key holders, with a stated but unevidenced intention to broaden it. The founding team is documented as the current governor.
- Position on the spectrum: multisig-controlled, not DAO-governed. The multisigs have 6, 13, 6 and 9 signers on-chain, but no source confirms signer identities or how many are outside the team.
- Roadmap: BENQI docs say the protocol will "progressively be decentralized and eventually be governed by a DAO" . BENQI also says QI distribution is meant to balance growth "while gradually decentralizing governance over time" . That is a stated goal, not a mechanism.
- Key-person risk: compromise of 4 of 6 upgrade-Safe signers could replace the implementation instantly. Compromise of the single withdraw key could move pooled AVAX. Assessment: a halt (pause) needs only 2 signers, so a rogue or compromised pause group could freeze minting and transfers.
- Token distribution: voting power over upgrades or parameters was not found, so token distribution cannot be matched to control. Miles voting power over validator delegation was not measured.
Governance Risks
- Single points of failure: the code-less
withdrawholder; the code-less owner of the reward contract'sProxyAdmin; and the 4-of-6 upgrade Safe with unnamed signers. - No exit protection: no timelock against a 15-day exit cooldown.
- Governance attacks: no vote on upgrades or parameters was found, so flash-loan or bribery attacks on such a vote are not established. Miles voting could influence which validators receive delegations; its concentration was not measured. Phishing or key theft of signers is the main vector.
- Centralization creep: the pause Safe also owns the lending
PauseGuardian, so one 2-of-13 group spans lending and staking emergencies. - Mitigations to credit: the Safes are standard, have no extra modules, and a wider 2-of-13 pause path exists besides the upgrade Safe. Sourcify verification and an Immunefi bounty exist (scope listed on Immunefi), which are assurance, not governance.
Data Completeness
- Available: on-chain role holders, Safe thresholds and owners, proxy admin chain, cooldown values (block 98915447); BENQI risk disclosures; a 2022 forum post on the multisig history; web pages from BENQI docs on Node Voting, Miles and the risk page.
- Missing: signer identities for all Safes; the MPC threshold behind the withdraw address; any timelock or change policy; vote data (the venue list is empty, and web searches found no Snapshot space or Governor); the on-chain per-role admin and the grant/revoke delay (inferred from source only).
- The audit files have unverified publication provenance, so this report uses them only through digests; none state a timelock or signer set.
Score: 1.6/10
Rubric security-score-v2.6; data completeness 0.75
- Admin control
- 1.00 points
- Change delay
- 1.00 points
- Decision process
- 3.50 points
| Factor | Level | Evidence | Weight | Points |
|---|---|---|---|---|
| Admin control | L0 | supported | 0.45 | 1.00 |
| Change delay | L0 | supported | 0.30 | 1.00 |
| Decision process | — | not disclosed | 0.25 | 3.50 |
32 citations · click markers to view evidence
Cross-Analysis
Cross-Risk Analysis: Benqi Staked Avax
Conclusion. Benqi Staked Avax (sAVAX, the liquid staking receipt token on Avalanche) has no recorded hack, but its main risks come from combining findings across reports. Instant admin powers, a single-key withdraw role, a 15-day exit that admins can lengthen, and unverifiable off-chain custody all point at the same weakness: holders must trust a few keys and operators. The audits cover the contract's logic better than they cover these trust assumptions. Confidence is moderate to good on the control facts, which were read on-chain, and weak on liquidity depth and holder concentration.
Scope: the DefiLlama record covers one product, the sAVAX staking contract on Avalanche, with TVL of about USD 200.5M . BENQI lending markets, Ignite and QI are other products. They matter here only where they share keys or hold sAVAX.
Top Risks for a Holder
-
Key or signer compromise at the custody and upgrade layer (affects all
sAVAXholders).- What can go wrong: one code-less address (a single key, or possibly an MPC-controlled wallet) holds
withdraw(uint256)with a 0-second delay . A 4-of-6 Safe can upgrade the contract with a 0-second delay . The 2022 Certora report says admins can withdraw "without limitation" (audit published on a host not verified as auditor-controlled). The same 4-of-6 Safe is the default admin, and the governance report assesses that it can grant and revoke every role includingROLE_WITHDRAW. That gives the Safe signers a second route to the withdraw function; the delay on role changes was not measured. The Certora trust finding is rated High (estimated) because the audit states no rating, and the governance report scores admin control and change delay at the lowest level . - Magnitude: a withdraw-key loss exposes the AVAX sitting in the contract at that moment, which was not measured . An upgrade-Safe compromise reaches the same liquid balance plus later deposits, because the pooled-AVAX cap is effectively unlimited, so exposure grows with deposits. Pooled AVAX is delegated to P-Chain validators through the MPC system, but the delegated share is not quantified , so the liquid and delegated balances are both unknown. The sources do not show that an upgrade alone reaches delegated AVAX. The total claim on the pool at the read block is 24,097,137 AVAX, about USD 200.5M of TVL in the current snapshot . That snapshot bounds today's pool, not future deposits. The USD 500,000 bounty is about 0.25% of TVL and excludes centralization and privileged-address impacts .
- Likelihood: low. Basis: no hack is recorded , and the Safe needs 4 signers. Nothing in the sources limits the damage if it does happen. The missing fact that would change the band is the withdraw key's custody: MPC-controlled would lower it, a plain hot key would raise it to medium.
- What can go wrong: one code-less address (a single key, or possibly an MPC-controlled wallet) holds
-
Frozen or slowed exit combined with a depeg and a lending-market cascade (affects
sAVAXholders and anyone usingsAVAXas collateral).- What can go wrong: direct redemption takes a 15-day cooldown plus a 2-day window, and the 4-of-6 Safe can change both with 0 delay . A 2-of-13 Safe can also pause . The risk page says that if too many MPC nodes are down, staking, unstaking and cross-chain transfers will be paused until the network is restored, and the sources give no time bound. That halt is an off-chain operational pause, not a documented pause of the contract's redemption function. An outage could impair exits only if it also stops the off-chain movement of AVAX that funds redemptions, which the sources do not show . The fast exit is an AMM swap, and the two
sAVAX/WAVAX pools shown hold USD 101.7K and USD 496.6K of liquidity (source reliability: unverified publisher). Two lending-market contracts hold about 78% of supply (source reliability: unverified publisher): . One is labelled as the BENQIsAVAXmarket and the other as the Aave AvalanchesAVAXmarket, so Aave is an external market with its own controls. If a discount or pause occurred, borrowers in those markets could be liquidated into AMM pools whose depth is unconfirmed, depending on their positions, collateral factors and market rules, which the sources do not give. The sources describe this as a plausible scenario from the structure, not an observed event . Both the pool depth and the holder shares are unconfirmed. - Magnitude: temporary loss of liquidity for the whole pool. Realised discount losses fall on sellers and liquidated borrowers; the Snowtrace page (unconfirmed) values the two lending contracts'
sAVAXat about USD 165M. No depeg history is measured, so the size of a discount is unknown. - Likelihood: Assessment: low to medium for a temporary discount in a sharp AVAX sell-off. The basis is the 15-day exit confirmed on-chain: holders who need cash sooner must sell on AMMs. Whether a discount follows depends on AMM depth, which is unconfirmed because the pool figures come from an unverified publisher. No depeg or liquidation event is measured. Low for a deliberate multi-week freeze, since it needs signer misconduct. Missing facts that would change the band: depeg history, current DEX depth (the depth dataset has no assets), borrower positions and a non-weak holder breakdown.
- What can go wrong: direct redemption takes a 15-day cooldown plus a 2-day window, and the 4-of-6 Safe can change both with 0 delay . A 2-of-13 Safe can also pause . The risk page says that if too many MPC nodes are down, staking, unstaking and cross-chain transfers will be paused until the network is restored, and the sources give no time bound. That halt is an off-chain operational pause, not a documented pause of the contract's redemption function. An outage could impair exits only if it also stops the off-chain movement of AVAX that funds redemptions, which the sources do not show . The fast exit is an AMM swap, and the two
-
Backing and exchange rate that holders cannot verify (affects all
sAVAXholders).- What can go wrong: the AVAX per
sAVAXrate (about 1.29) is raised when a privileged role sends rewards toaccrueRewards(). In the public repository code, redemption also subtracts the paid AVAX fromtotalPooledAvax, and that code was not compared with the deployed version. The contract does not read P-Chain balances . The economic report finds no slashing rule, insurance fund or independent proof of reserves . If delegated AVAX is lost (validator slashing, MPC failure, operator error), the rate can stay unchanged until holders discover the gap at redemption. Assessment: the code review gives little comfort here, because the audited revision is not tied to the deployed implementation , and the audit coverage of the 955 critical-rated lines is unknown because the comparison was not performed. - Magnitude: an unknown share of pooled AVAX, because the delegated share is not quantified.
- Likelihood: low. Basis: no loss event in the records, and the project documents MPC controls. The missing facts are the MPC threshold and signer set, the validator slashing handling, and a deployed-versus-audited code comparison.
- What can go wrong: the AVAX per
Combined Attack Vectors
Vector 1: Withdraw-key compromise with an unchanged exchange rate
- An attacker obtains the key behind
0x565f…7e83(phishing, host compromise, or an MPC threshold breach). - The attacker calls
withdrawwith no amount limit and no delay, sending contract AVAX to the caller. The code shows only a role check . - In the older docs copy of the code the rate is internal accounting (
totalPooledAvaxover shares), so the quoted value ofsAVAXwould stay near 1.29 AVAX while the liquid balance falls . The deployed update rule is not established from citable text, so this step is limited to the documented code and has not been tested. - If the contract's liquid AVAX is insufficient, redemption transfers would fail after the 15-day cooldown. The older docs copy of the code requires the AVAX transfer to succeed , and it is not confirmed to match the deployed version. If a discount appeared, it would show first in AMM prices; no such discount is measured.
- The same outcome is reachable by compromising the 4-of-6 default-admin Safe, which can reassign
ROLE_WITHDRAW(the delay on role changes was not measured). - The bounty does not pay for this path, because privileged-address attacks are out of scope.
- Domains: code (unrestricted role, audit-accepted trust assumption), governance (single key, 0 delay), economics (no documented loss cover, no reserve proof), social (signers and key custody not named).
- Likelihood: low. Basis: needs the key, and the custody setup is undisclosed.
- Impact: severe, up to the contract's liquid AVAX at that time.
Vector 2: Upgrade-Safe takeover inside the exit window
- Four of six Safe signers are compromised or collude. The on-chain owner lists show the same six addresses own both the 4-of-6 upgrade Safe and the 3-of-6 Pause Guardian Safe, so one signer group holds upgrade, resume and guardian pause rights. No source names them .
- The Safe upgrades the
StakedAvaxproxy at once. A rogue implementation can change balance logic . - A holder who has not yet completed the 15-day cooldown cannot redeem through the protocol before an instant upgrade takes effect, and their only faster exit is an AMM sale. A holder who has already served the cooldown and is inside the 2-day redeem window could still redeem before the upgrade, if the contract holds enough liquid AVAX. The size of that group is not measured. A paused contract does not stop an upgrade.
- Any pause by the 2-of-13 Safe cannot undo it, because
resume()sits with the 4-of-6 Safe.
- Domains: governance (0 delay), economics (15-day exit), code (deployed version not compared with audits), social (anonymous signers).
- Likelihood: low. Basis: a 4-of-6 threshold, and no history of misuse.
- Impact: severe. The upgrade reaches the liquid contract balance and all later deposits and can block exits; delegated P-Chain AVAX is not shown to be reachable by an upgrade alone. Today's pool is about USD 200.5M of TVL (current snapshot); later deposits would add to the exposure.
Vector 3: Stress cascade through shared pause keys and lending markets
- AVAX falls sharply. USD TVL already moves with the AVAX price .
- Holders rush to AMM pools whose depth is unconfirmed.
sAVAXtrades below its redemption value, and, if borrower positions are close to their limits, borrowers in the two lending markets that hold about 78% of supply (weak source) could be liquidated. - Operators can use emergency functions, and an MPC outage halts staking and unstaking off-chain. Whether either stops the contract's redemption function is not established. The risk page warns that emergency functions "could affect your ability to withdraw assets or redeem sAVAX" .
- The same 2-of-13 Safe can pause
sAVAX, and it owns the BENQI lending-marketPauseGuardian, wherepauseTransfers()is a measured privilege with a 0-second delay . The measured facts are that shared ownership and that privilege; the guardian's own read showscanPause: falseandareAllMarketsPaused: truefor the BENQI Comptroller . Whether one decision could freeze the BENQIsAVAXmarket depends on the guardian's actual powers and market coverage, which are not established. The Aave market is outside this shared control.
- Domains: economics (liquidity, exit), governance (shared pause key, no bounds), dependency (MPC, lending markets), code (pause semantics).
- Likelihood: low to medium for the discount, lower for the liquidation part (borrower positions and market rules are unknown), low for a prolonged freeze.
- Impact: moderate for plain holders who can wait out the cooldown; severe for leveraged users only if liquidations occur.
Systemic Risk Factors
- Protocol dependencies: the dependency graph lists no protocol backing the TVL, only Avalanche at share 1.0 and no warnings . Composition is missing, so this is an absence of data, not a clean result. The real dependencies are the off-chain MPC signer set, and the validators. One document also describes a staking bot that reduces delegated AVAX for redemptions, which is unconfirmed (source reliability: unverified publisher). They are routed to the owner of the dependency-risk factor, which scored it at the lowest level .
- Chain-level: Avalanche is a base chain, not a rollup, with no L2 sequencer risk . Both the contract and the P-Chain staking depend on the same network. A chain-wide halt would stop both the exit and any AMM escape.
- Macro:
sAVAXis a receipt for AVAX, so a market-wide sell-off hits price, collateral values and AMM depth together. - Correlation: the BENQI lending market that holds about 40% of supply (weak source) has a pause guardian owned by the same 2-of-13 Safe that can pause
sAVAX, and the Aave market that holds about 37% is external. Weak-tier sources report that a sibling BENQI market was paused in May 2026 after an oracle-provider wallet compromise; this is unconfirmed, and the project says no user funds were impacted . The report is not evidence of ansAVAXincident or of an exposure of this protocol.
Risk Amplifiers
- Overlapping signers: the on-chain owner lists of the 4-of-6 upgrade Safe and the 3-of-6 Pause Guardian Safe are the same six addresses, and all six also appear among the owners of the 2-of-13 pause Safe and the 3-of-9 protocol-fee Safe . Assessment: upgrade, resume, guardian pause and fee control are not independent, so one group of six keys is the real trust base.
- Instant powers + unnamed signers: a 0-second upgrade delay inside a 15-day exit, with signers no source names. This is what leaves holders who have not yet served the cooldown without a protocol exit after a signer compromise.
- Single-key withdraw + no reserve proof + no documented loss cover: if the key fails, no fund or proof is evidenced to catch the loss; its size, trigger and payout rules are unknown.
- Audit gaps + upgradeability: audits date from 2021 to 2022 or carry no stated year, the deployed implementation was not compared with them, and an instant upgrade can invalidate them anyway. Seven of nine audit PDFs sit on a host not verified as auditor-controlled. This lowers confidence in them, not the score.
- Possible concentrated supply + limited AMM depth: if the pool and holder figures from unverified publishers hold, the AMM exit is shallow compared with the holder base. Neither is confirmed.
- Rewards contract named a timelock but upgradeable by a single code-less address with 0 delay: the
ROLE_ACCRUE_REWARDSrole that feeds the rate is held by two addresses, this contract and the 4-of-6 Safe . The contract's upgrade control (a single code-less address, 0 delay) is weaker than the Safe's, and the contract's own delay was not measured .
Risk Mitigators
- Two firms plus formal verification on
StakedAvax: Halborn's four critical findings are marked solved, and Certora reported no major issue in the contract (both on a host not verified as auditor-controlled) . - Provider-reported runtime match: Sourcify reports a runtime match for the proxy. This is not independent assurance of the code or of audit-to-deployment equivalence .
- Wider pause path (partial): any two of the 2-of-13 Safe's 13 owners can pause, and seven of those owners are not on the upgrade Safe. The guardian's owners are the same six as the upgrade Safe, so the pause paths are not independent of upgrade signers.
- No custody of QI or leverage in the contract: the contract has no borrowing, and QI does not back
sAVAX. - Track record: no hack recorded and live since 2022-02-17 . The social report scores incident record at the top level but resilience as not disclosed.
- Bug bounty: up to USD 500,000 for critical smart-contract bugs, with the reward timelock address in scope .
- Aligned revenue: the protocol takes 10% of rewards , so its income depends on TVL.
Contradictions & Inconsistencies
- Withdraw key custody: governance says the address is "probably the MPC-controlled operator wallet" . The economic report says it is not documented whether it is MPC-controlled, and rates the attack as "low to moderate" . The code report calls it a single key. The only support for MPC is a 2022 team post, so this report treats it as unresolved. It does not lower the risk.
- Certora Aave adapter audit date: the code report and overview say the year is not stated, while the economic report calls it "the 2022 Certora review". The audit file carries no date , but BENQI's risk page lists it as June 2022 . The economic report's 2022 date is supported by that official listing.
- Oracle: this is a scope split, not a conflict. The economic report separates the internal mint and redeem rate from downstream lending feeds that belong to other products . It scores oracle design as not disclosed because the deployed update-size limit is undocumented .
- Exchange-rate mechanics: the economic report says the rate is changed only by holders of the rewards role . The public repository code also lowers
totalPooledAvaxon redemption (see Top Risk 3). The repository version was not compared with the deployed contract, so the size of this effect is unknown. - Source copy vs deployed code: the economic report treats the docs copy of
StakedAvax.solas older than the deployed contract, becauseaccrueRewards(uint amount)differs from the deployedaccrueRewards(). The overview and governance reports still cite the docs copy for some mechanics. Statements built on the docs copy are not confirmed against the deployed version. - Pause Safe ownership: the overview says the Pause Guardian
0x10f3…is owned by a separate 3-of-6 Safe. Governance says the 2-of-13 Safe owns the lendingPauseGuardian, a different contract (0xaad2…). Both can be true. The on-chain owner lists show the Pause Guardian Safe (3-of-6) and the upgrade Safe (4-of-6) have identical owners, and the 2-of-13 Safe contains all six. - Reward contract name: the bounty calls it a "timelock", while on-chain control shows an EOA-owned upgradeable proxy with 0 delay. The name does not match the measured control.
- Scores: governance 1.6/10, economic 3.8/10, social 4.0/10 and code 6.0/10 are consistent with each other. The track-record score of 8.6/10 sits well above them, because it measures the absence of recorded incidents, not control quality.
Overall Risk Profile
Benqi Staked Avax is a simple, mature, single-chain liquid staking pool with an admin-operated custody layer. Its weak scores come from governance and economics, not from known coding flaws. No sAVAX incident is recorded, so the most likely failure mode cannot be ranked with confidence. The structural candidate is a stress-time exit squeeze: the 15-day exit, confirmed on-chain, pushes sellers to AMMs whose depth is unconfirmed, and a pause or longer cooldown would worsen it. Whether lending liquidations follow depends on borrower positions and market rules, which are not established. The worst case is an instant upgrade or withdraw-key compromise that reaches pooled AVAX, with no timelock, no reserve proof and no documented loss cover. For comparison with other liquid staking tokens, the sources contain no peer data, so no comparison is made. Against the usual expectation of a timelock on upgrades, the 0-second delay is a weaker setup.
- BENQI sAVAX market
- 40.6 percent
- Aave Avalanche SAVAX
- 37.4 percent
Data Gaps
- Some audit reports were unavailable or could not be read in full.
- An audit report is unavailable because it could not be collected.
- Current deployment addresses could not be confirmed for some audited contracts.
- The liquid AVAX balance in the
StakedAvaxcontract at the read block is not measured, so the Vector 1 loss cannot be sized. - The custody and signing setup of the withdraw address, the MPC threshold and the signer sets of all Safes are not disclosed. The people behind the shared six owner addresses are not named.
- Per-role admin (
grantRole,revokeRole) and its delay were not read on-chain. The governance report assesses from source code that the 4-of-6 Safe can reassignROLE_WITHDRAW; this is not confirmed on-chain. - What
areAllMarketsPaused: trueon the BENQI lendingPauseGuardianmeans for the BENQIsAVAXmarket is not established. - Current DEX depth is a gap (the dataset has no assets). The two GeckoTerminal pools and the Snowtrace holder list come from unverified publishers.
- The deployed implementation was not compared with any audited version, and the audited commits are not public. The share of pooled AVAX delegated, the slashing handling, and the design of any staking bot are not documented in a confirmed source.
- Aave listing state and the exposure of other integrators were not established, so the size of the liquidation cascade in Vector 3 cannot be computed.
- The May 2026 sibling-market event has no formal post-mortem and only weak-source coverage. Whether it touched
sAVAXis not established.
56 citations · click markers to view evidence